臺灣博碩士論文加值系統

目的 : 使用人工智慧語義辨識的方法輔助皮膚科症狀的診斷，驗證詞向量結合醫療知識圖譜應用於醫療輔助診斷是簡單有效的語義辨識方法。
材料與方法 : 本研究首先建立皮膚科症狀與疾病的知識圖譜，在醫療症狀的描述句子是一個症狀的關鍵詞就是一個句子，並賦與相應的發生部位，再建立此症狀可能發生疾病的關聯與機率。第二是建立詞向量 (word embedding)及處理使用者的輸入句子是用自然語言處理方法 (Nature Language Process, NLP)，使用開放程式庫Jieba的分詞法 (Word Segmentation)及去除一些停用詞 (Stop Words)，並運用Thomas Mikolov提供的Word2vec公用程式機器學習WiKi大量的文章，建立漢語通用語詞的詞向量來比對使用者輸入的症狀句子與醫療知識圖譜的症狀描述語句的相似度，排序出最可能相似的症狀描述再針對症狀的發生部位做篩選增進判斷的準確率，達到語義辨識輔助皮膚科診斷的應用。
結果 : 本研究開發一個應用程式：建立皮膚科症狀與疾病的關聯資料庫、詞向量的相似度計算，提供使用者輸入皮膚科的症狀即可顯示相關疾病的百分比。並針對皮膚科的六種疾病做比對，詞向量的訓練模型目標單詞模型 (Skip-gram)誤差範圍最大是0.228、最小是0.086，CBOW(Continuous Bag of Word)的誤差範圍最大是0.617、最小0.138，所以在醫療句子的語義辨識上，目標單詞模型 (Skip-gram)模型優於CBOW模型。
結論 : 本研究提出一個非常有效而簡單的語義辨識應用於輔助皮膚科的症狀診斷，這個方法很容易擴充至不同的醫療科別甚至各個專業領域，只要建立好專業領域的知識圖譜結合訓練好的詞向量，即可有效的辨識句子的語義。

Purpose: Semantic recognition based on artificial intelligent method to assist the diagnosis of dermatological symptoms. To verify word vector combined with the knowledge map of medical aided diagnosis is a simple and effective semantic recognition method.
Materials and method: This study first establishes a knowledge map of dermatological symptoms and diseases. The description of the medical symptoms is a sentence with only one keyword, and assigned to the corresponding site of occurrence, and then establish the association and probability of this disease may occur. The second is to create a word embedding by using the open library Jieba Word Segmentation and removing some stop words, and the Word2vec utility provided by Thomas Mikolov to learn huge WiKi articles. After establishing the word vector of Chinese common words, to compare the similarity between the symptom sentences input by the user and the symptom description statements of the medical knowledge map. The most likely similar symptom description with its corresponding site of occurrence is used to improve the accuracy of the judgment, and to achieve the application of semantic recognition to assist the diagnosis of dermatology.
Results: This study developed an application that establishes a database of associations between dermatological symptoms and diseases, and a similarity calculation for word vectors, providing users with the input of dermatological symptoms to show the percentage of related diseases. As well, it is aimed at the six diseases of dermatology to verify the training models of the word vector which is suitable in the semantic recognition of medical symptom sentences. The Skip-gram model has a maximum error of 0.228 and a minimum of 0.086. The maximum error of CBOW (Continuous Bag of Word) model is 0.617 and the minimum is 0.138, so the Skip-gram model is superior to the CBOW model in the semantic recognition of medical sentences.
Conclusions: This study proposes a very effective and simple semantic recognition applied to assist in the diagnosis of dermatological symptoms. Also, this method can be easily extended to different medical departments and even various professional fields. As long as the knowledge map of the professional field and the trained word vector are established. The semantics of the sentence can be effectively recognized.

摘要 I
ABSTRACT II
致謝 IV
目錄 V
表目錄 VI
圖目錄 VII
符號縮寫 VIII
符號說明 IX
第一章 緒論 1
1.1 動機 1
1.2 目的 2
1.3 相關文獻探討 2
第二章 材料與方法 5
2.1 前言 5
2.2 醫學知識圖譜的建立 6
2.2.1 資料庫的維護 6
2.2.2 醫學書籍及康健醫療網 7
2.2.3 關聯圖的建立 8
2.3 詞向量的建立 9
2.3.1 中文分詞工具 9
2.3.2 大量中文詞句的收集 10
2.3.3 詞向量 12
2.4 症狀詞義辨識 15
2.4.1 中文詞句的語義相似度 15
2.4.2 語義相似度結合症狀部位比對 16
第三章 結果 17
3.1 前言 17
3.2 醫療關聯圖 18
3.2.1 醫療資料庫維護 18
3.2.2 症狀與疾病關聯圖 19
3.2.3 疾病圖示與診療參考 20
3.3 皮膚科診斷應用例 20
3.3.1 香港腳、足癬 21
3.3.2 褥瘡 23
3.3.3 帶狀皰疹 25
3.3.4 膠原病 27
3.3.5 異位性皮膚炎 29
3.3.6 表皮囊腫 31
3.4 詞向量不同訓練模型的結果 33
第四章 討論 34
第五章 結論 36
參考文獻 37
自傳 40

[1]Heaven, D., Your next doctor's appointment might be with an AI, in MIT technology review. 2018.
[2]Hirschberg, J. and C.D. Manning, Advances in natural language processing. Science, 2015. 349(6245): p. 261-266.
[3]Newman, D., et al., Distributed algorithms for topic models. Journal of Machine Learning Research, 2009. 10(Aug): p. 1801-1828.
[4]Deray, T. and P. Verheyden. Towards a semantic integration of medical relational databases by using ontologies: a case study. in OTM Confederated International Conferences" On the Move to Meaningful Internet Systems". 2003. Springer.
[5]Velupillai, S., et al., Recent advances in clinical natural language processing in support of semantic analysis. Yearbook of medical informatics, 2015. 24(01): p. 183-193.
[6]Pons, E., et al., Natural language processing in radiology: a systematic review. Radiology, 2016. 279(2): p. 329-343.
[7]Harispe, S., et al., Semantic similarity from natural language and ontology analysis. Synthesis Lectures on Human Language Technologies, 2015. 8(1): p. 1-254.
[8]Todhunter, J., et al., Semantic processor for recognition of cause-effect relations in natural language documents. 2015, Google Patents.
[9]Hebert, M., et al., Multi-domain natural language processing architecture. 2019, Google Patents.
[10]鄭捷, 自然語言處理：用人工智慧看懂中文. 2018, Taipei, ROC: 洪錦魁.
[11]Li, B., et al., Chinese Sentence Similarity Computing Based on Semantic Dependency Relationship Analysis [J]. Application Research of Computers, 2003. 12: p. 15-17.
[12]フォッツィ著，陳韻如譯, 人體疾病大解密. 2015, New Taipei, ROC: 楓書坊文化出版社.
[13]Budd, J.R., et al., A medical information relational database system (MIRDS). Computers and Biomedical Research, 1988. 21(5): p. 419-433.
[14]呂耀卿, 中國人皮膚病圖譜(Skin Diseases in Chinese). Vol. 1,2,3. 1995.
[15]Delgado, M., et al., Mining association rules with improved semantics in medical databases. Artificial Intelligence in Medicine, 2001. 21(1-3): p. 241-245.
[16]Yang, S.-H., et al. Large-scale high-precision topic modeling on twitter. in Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 2014. ACM.
[17]Pennington, J., R. Socher, and C. Manning. Glove: Global vectors for word representation. in Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP). 2014.
[18]Kiros, R., et al. Skip-thought vectors. in Advances in neural information processing systems. 2015.
[19]Mikolov, T., et al. Distributed representations of words and phrases and their compositionality. in Advances in neural information processing systems. 2013.
[20]Mikolov, T., et al., Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013.
[21]Grave, E., et al., Learning word vectors for 157 languages. arXiv preprint arXiv:1802.06893, 2018.
[22]Levy, O., Y. Goldberg, and I. Dagan, Improving distributional similarity with lessons learned from word embedding. Transactions of the Association for Computational Linguistics, 2015. 3: p. 211-225.
[23]Mao, X., et al. Chinese word segmentation and named entity recognition based on conditional random fields. in Proceedings of the Sixth SIGHAN Workshop on Chinese Language Processing. 2008.
[24]Uijlings, J.R., et al., Selective search for object recognition. International journal of computer vision, 2013. 104(2): p. 154-171.
[25]Zhang, Q., X. Liu, and J. Fu. Neural networks incorporating dictionaries for Chinese word segmentation. in Thirty-Second AAAI Conference on Artificial Intelligence. 2018.
[26]Lott, B., Survey of keyword extraction techniques. UNM Education, 2012. 50.
[27]Sun, J., ‘Jieba’ Chinese word segmentation tool. 2012.
[28]Levow, G.-A. The third international Chinese language processing bakeoff: Word segmentation and named entity recognition. in Proceedings of the Fifth SIGHAN Workshop on Chinese Language Processing. 2006.
[29]Zhang, X. and Y. LeCun, Text understanding from scratch. arXiv preprint arXiv:1502.01710, 2015.
[30]Yao, L., D. Mimno, and A. McCallum. Efficient methods for topic model inference on streaming document collections. in Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining. 2009. ACM.
[31]Hinton, G.E. Learning distributed representations of concepts. in Proceedings of the eighth annual conference of the cognitive science society. 1986. Amherst, MA.
[32]Hinton, G.E. and R.R. Salakhutdinov, Reducing the dimensionality of data with neural networks. Science, 2006. 313(5786): p. 504-507.
[33]Le, Q. and T. Mikolov. Distributed representations of sentences and documents. in International conference on machine learning. 2014.
[34]Cho, K., et al., Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078, 2014.
[35]Molino, P. and L.M. Aiello. Distributed Representations for Semantic Matching in non-factoid Question Answering. in SMIR@ SIGIR. 2014.
[36]Elkan, C. and K. Noto. Learning classifiers from only positive and unlabeled data. in Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining. 2008. ACM.
[37]Mikolov, T., Q.V. Le, and I. Sutskever, Exploiting similarities among languages for machine translation. arXiv preprint arXiv:1309.4168, 2013.
[38]Mikolov, T., Statistical language models based on neural networks. Presentation at Google, Mountain View, 2nd April, 2012. 80.
[39]Jiu-le, T. and Z. Wei, Words Similarity Algorithm Based on Tongyici Cilin in Semantic Web Adaptive Learning System [J]. Journal of Jilin University (Information Science Edition), 2010. 6(010).
[40]Russell, S.J. and P. Norvig, Artificial intelligence: a modern approach. 2016: Malaysia; Pearson Education Limited.
[41]Cambria, E., Affective computing and sentiment analysis. IEEE Intelligent Systems, 2016. 31(2): p. 102-107.
[42]Schmidhuber, J., Deep learning in neural networks: An overview. Neural networks, 2015. 61: p. 85-117.
[43]LeCun, Y., et al., Gradient-based learning applied to document recognition. Proceedings of the IEEE, 1998. 86(11): p. 2278-2324.
[44]Srivastava, N., R.R. Salakhutdinov, and G.E. Hinton, Modeling documents with deep Boltzmann machines. arXiv preprint arXiv:1309.6865, 2013.
[45]Sundermeyer, M., R. Schlüter, and H. Ney. LSTM neural networks for language modeling. in Thirteenth annual conference of the international speech communication association. 2012.
[46]程鼎元, et al., 智慧型觀光醫療服務推薦系統之設計與實作. 資訊管理學報, 2009. 16(_S): p. 131-152.
[47]劉知遠, 基于文檔主題結構的關鍵詞抽取方法研究. 2011, 北京：清華大學.
[48]Heinrich, G., Parameter estimation for text analysis. 2005, Technical report.
[49]Kim, Y., Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882, 2014.
[50]Moghaddam, S. and M. Ester. On the design of LDA models for aspect-based opinion mining. in Proceedings of the 21st ACM international conference on Information and knowledge management. 2012. ACM.
[51]Girshick, R., et al. Rich feature hierarchies for accurate object detection and semantic segmentation. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2014.